driver telematics analysis for kaggle competition containing all processing performed after February 20, 2015. Much of the code in this repository was built on a template provided by [Stephane Soulier] (http://www.kaggle.com/c/axa-driver-telematics-analysis/forums/t/11299/score-0-66-with-logistic-regression) in the Kaggle Competition Forumn.
Brief Summary of Code
- gbm_speedhistogram_tripLength_wProcessedDat.R - gradient boosting model using doParallel for parallel backend, features: trip length and speed histogram - score: 0.78
- gbm_speedhistogram_doMC.R - gradient boosting model using doMC for parallel backend - score: 0.78
- svm_speedhistogram_tripLength_wProcessedDat.R - svm using speed histogram and trip length - score: 0.73
- gbm_speedhistogram_removeOverlap.R - gradient boosting model (not parallel), filter out target 0 data from targe 1 entries - score: 0.69
- h20_deepLearning.R - uses deep learning from h2o package - score: 0.61
- gbm_speedquantile.R - gradient boosting model using doMC backend - score: 0.77
- prediction_triplength_glm.R - logistic regression using previously compiled stop data - score: 0.56